Haze

via Instagram https://instagr.am/p/DTrsJucgu8v/

Gloves

via Instagram https://instagr.am/p/DTqKVEXgiYd/

I just posted this on Twitter as @ rvanbruggen

This was the text:

Crossposting this from #bsky: I just wrote something about AI-shaped holes, and would love for you to poke holes AT IT! https://t.co/KPZUWhxDeH. See https://t.co/0YO8eSyBoN for the original.

Crossposting this from #bsky: I just wrote something about AI-shaped holes, and would love for you to poke holes AT IT! https://t.co/KPZUWhxDeH. See https://t.co/0YO8eSyBoN for the original.

— Rik Van Bruggen (@rvanbruggen) Jan 15, 2026

from Twitter https://twitter.com/rvanbruggen

January 15, 2026 at 11:28AM
via IFTTT

Stop looking for AI-shaped holes!

Why enterprises need to reimagine what’s possible

If you talk to enterprise leaders about artificial intelligence today, a familiar pattern emerges. Most conversations start with a very concrete question: Which problem should we solve with AI first?

That question feels sensible. Enterprises are built around prioritization, ROI, and risk management. But when it comes to generative AI, this framing can also be a trap. Starting with narrowly defined problems often leads to narrowly defined outcomes – and that’s not where the real value of this technology lies.

Generative AI is not just another optimization tool. It’s a new way of designing work, coordination, and decision-making. And enterprises that treat it as a solution in search of a predefined problem risk missing much larger opportunities for innovation.

The limits of “problem-first” AI thinking

In many organizations, AI initiatives begin by looking at an existing workflow and asking whether AI can make it faster or cheaper. Reduce average handling time. Deflect a percentage of tickets. Summarize documents more efficiently.

All of that is useful – but it assumes the underlying process is fundamentally sound.

History suggests that transformational technologies rarely deliver their full value that way. Email didn’t just speed up memos. Cloud computing didn’t just reduce data center costs. Smartphones didn’t simply digitize paper workflows. Each of these technologies changed how work was structured in the first place.

Generative AI belongs in that same category. Its real impact comes not from incremental improvement, but from rethinking how work flows across people, systems, and data.

So what can we do differently? What specific steps could we take to NOT look for the AI-shaped hole, but to truly use this technology at its full potential?

Here are a few practical shifts to consider.

1. Change the mechanics

For many enterprise stakeholders, “generative AI” still means one thing: a chatbot that looks and behaves like ChatGPT. You ask a question, it responds with text, and you iterate in real time.

That model is familiar – but it is only one possible interaction pattern, and often not the most effective one in an enterprise setting.

Much of enterprise work is asynchronous. Requests arrive through email, internal portals, or messaging tools. Context accumulates over time. Decisions are rarely made in a single back-and-forth interaction. I have written about this in the past – and stand by it.


The distinction between synchronous and asynchronous communication media opens the door to a different kind of innovation: changing the mechanics of interaction. AI agents that operate via email, messaging platforms, or background workflows often align far better with how work actually happens. In many cases, the AI doesn’t need a conversation at all – it needs access to context, ownership, and the ability to act.


Simply changing the communication medium can already unlock major gains in adoption, efficiency, and user satisfaction—without changing the underlying AI model.

2. Taking the process helicopter view

To go further, enterprises need to zoom out. Instead of examining AI opportunities task by task, it helps to take a “process helicopter” view of the organization.

From 50,000 feet up, two questions become especially revealing.

• Which processes are currently performed by humans, are repetitive in nature, and consume a significant amount of time.
• Which processes are performed by humans, deal primarily with unstructured information – emails, documents, chat messages, requests – and also consume a lot of time.

These questions cut across roles and departments. They surface work that exists not because it creates value, but because humans have historically been the only way to connect systems, interpret context, and move work forward.

With the current state of AI technology, many of these process steps are highly automatable. More importantly, they are often redesignable. Platforms like DevRev illustrate this by treating conversations, work items, and systems of record as part of a single operational fabric, rather than separate silos.

And tools like DevRev’s Agent Studio take the grunt out of designing these processes – and make it super easy to “land” the helicopter, and improve your processes.

What this looks like in practice

To make this more concrete, consider a few common enterprise scenarios.

• In customer support organizations, a large amount of work happens before an issue is ever resolved. Tickets are triaged, clarified, routed, enriched with context from product and CRM systems, and handed off between teams. Traditionally, this coordination work is invisible but time-consuming. AI can take ownership of much of this orchestration – reading incoming conversations, creating and updating work items, linking them to the right customers and products, and escalating only when human judgment is truly required.
• In product organizations, feedback from customers often arrives as unstructured input scattered across emails, support tickets, call transcripts, and chat logs. Humans manually summarize this information and attempt to translate it into roadmap decisions. AI systems can continuously ingest these signals, cluster them by theme, and connect them directly to product work – shortening the distance between customer conversations and engineering action.
• In internal operations, think about onboarding, access requests, or policy questions. These are rarely complex, but they are highly repetitive and distributed across multiple systems. Instead of creating yet another portal or chatbot, AI agents can operate across existing channels, interpret intent, gather context, and execute actions across systems – while keeping humans in the loop only when exceptions arise.

In all of these cases, the real innovation is not “AI answering questions.” It’s AI owning work, understanding context, and moving processes forward end to end.

3. Structured ways to innovate with AI

As a last point, I would also like to remind you that thinking differently about AI requires more than inspiration – it requires method. There are lots of different methods out there that can structurally help you innovate, and I am sure every other professional will have their preference. But it’s pretty clear that you can take a deliberate approach, and strategically structure your thought process and come up with different ways to innovate in your organisation, using AI.

Let me give you a few examples of approaches you could take:

• One useful approach is Jobs To Be Done. Instead of focusing on tasks or tools, teams ask what job a process is truly trying to accomplish. AI often makes intermediate steps unnecessary, allowing entire workflows to collapse.
• Another powerful technique is process inversion. Teams map a workflow and then ask what would remain if humans were removed from it entirely. This quickly reveals which steps exist because of historical constraints rather than real value creation.
• A third approach is zero-based process design. Teams design workflows from scratch under the assumption that AI can read, understand, and act on unstructured information by default. Humans are then reintroduced intentionally, rather than by habit.

Finally, some organizations flip the question entirely by starting with AI capabilities instead of problems. They map what AI can reliably do today – classification, reasoning, summarization, decision support, autonomous action – and explore where those capabilities could enable entirely new operating models.

A different question to ask

Perhaps the most limiting question in enterprise AI strategy is: Where can AI help us with our current problems? Don’t do that - it only leads to self-limiting options! A far more powerful one is: If AI were native to our organization, what would we never design this way again?

Enterprises that ask that question won’t just automate faster. They will operate differently. And over time, that difference – not incremental optimization – is where lasting competitive advantage will be built.

At DevRev, we ask for nothing better than to work together on answering that intellectually challenging question. Let’s engage, and instead of solving problems, work together on unlocking opportunities!

Cheers

Rik

Frosty

via Instagram https://instagr.am/p/DTGrw3Wgg3z/

Circle

via Instagram https://instagr.am/p/DSzbyLgAof4/

Follow

via Instagram https://instagr.am/p/DSu8A2HAu_m/